3 user purchasing history. Meanwhile, the collaborative filtering is recommendation based on
other user’s preferences that similar to the active user likings. The demographic filtering concerned on the likings based on demographic similarity, for example age or gender. Lastly,
hybrid filtering is recommendation based on combining two filtering algorithms, for example combination between content-based filtering and collaborative filtering or
combination between demographic filtering and collaborative filtering. Another filtering algorithm is the rating-based item-to-item algorithms. According to Lemire Mcgrath
2013 this which is the slope-one algorithm involve in predicting the similarity differences in item rather than user similarity.
Nowadays, the current research focuses on frequently purchased products such as books, movies, and songs because this product are inexpensive thus it is frequently
purchased by consumer. Also, for frequently purchased product the accumulation of rating data and reviews are high in number. Thus, making recommendation for frequently
purchased product more efficient. Recently, researchers are exploiting knowledge-based recommendation techniques for infrequently purchased products. However, this technique is
more applicable in case-based reasoning, meaning it requires a profound knowledge of the product domain in making recommendation to users.
1.2 Problem statement
Due to tourism information overflow, tourists are being surrounds by invalid information that indirectly hinders the legitimate tourism information from being delivered.
This situation can be troublesome to tourists as traveling does cost a fair amount of money and a right decision can help them save up and thus helping them in making travel
expenditure beforehand. This study is motivated by several problems which are:
i. Knowledge-based technique in recommender system for infrequently purchased
products or items require extensive knowledge reengineering process. ii.
Most of rating data for infrequently purchased products are unavailable because tourists rarely stayed or rarely check-in to the hotel
iii. The current recommender system for infrequently purchased products or items
requires substantial user involvement. However, tourists are unwillingly to provide input such as reviews or ratings which are important in providing recommendations.
4
1.3 Research Questions
Based on the problem statement above, there are several research questions that stimulates this study, which are:
i. Which of the recommendation techniques can assist in recommending the
infrequently purchased products or items. For example, in tourism industry which hotel best fit the tourist interest and should be recommended?
ii. Which data from the information collected can be utilized to recommend
infrequently purchased products without requiring extensive users’ involvement? iii.
Which algorithm best suited with the chosen recommendation techniques to optimize recommendations of the infrequently purchased products or items.
1.4 Research Objective
Motivated by the problem statements and the research questions stated in the previous section, this study has three known objectives which are:
i. To study different types of recommendation techniques for infrequently purchased
products and to investigate technique and dataset that are suitable to implement in recommending infrequently purchased products.
ii. To developed a prototype in recommending hotels to user using user’s action view
and user rating data. iii.
To evaluate the proposed recommendation techniques using user testing evaluation.
1.5 Research Scope and Limitation
